AgentSkillsCN

brain-briefing

基于转录历史生成全面的项目简报,助力大脑快速上手与融入团队。

SKILL.md
--- frontmatter
name: brain-briefing
description: Generate comprehensive project briefings from transcript history for brain onboarding
category: autonomy

Brain Briefing

Purpose

Generate comprehensive briefings from transcript history to help autonomous brains get up to speed on project work, decisions, and current state.

Use Cases

  1. Brain Onboarding: New brain deployed to a project needs full context
  2. Cross-Directory Context: Understand work across parent and child directories
  3. Task-Transcript Correlation: See which sessions completed which tasks
  4. Team Coordination History: Understand agent team work and collaboration
  5. Decision Timeline: Reconstruct key decisions and their rationale

When to Use

Use brain-briefing when:

  • Deploying a new brain to a project
  • Resuming work after long absence
  • Understanding child project relationship to parent
  • Debugging: "what work led to current state?"
  • Knowledge transfer between brains

Usage

bash
# Basic briefing for a directory
/brain-briefing ~/signal

# Include child directories recursively
/brain-briefing ~/signal --recursive

# Filter by date range
/brain-briefing ~/signal --since "2026-02-01"
/brain-briefing ~/signal --last 7  # Last 7 days

# Focus on specific aspects
/brain-briefing ~/signal --tasks-only
/brain-briefing ~/signal --decisions-only

# Output formats
/brain-briefing ~/signal --format markdown  # Default
/brain-briefing ~/signal --format json      # Machine readable

How It Works

1. Transcript Discovery

  • Searches ~/.claude/projects/ for transcripts matching directory path
  • Optionally includes child directories (e.g., ~/signal/signal-workers-v2)
  • Sorts by timestamp (most recent first)

2. Task Correlation

  • Queries ~/.claude/tasks/ for task list
  • Correlates task IDs mentioned in transcripts with current task state
  • Builds task timeline showing progression across sessions

3. Team Analysis

  • Checks ~/.claude/teams/ for agent team configurations
  • Identifies teammate coordination and messaging patterns
  • Shows which teammates worked on which tasks

4. Briefing Generation

Produces structured output with:

  • Summary: High-level overview of project activity
  • Timeline: Chronological work history
  • Tasks: Current state, completed work, blockers
  • Teams: Agent team coordination (if applicable)
  • Key Decisions: Important choices and rationale
  • Files: Most frequently modified files
  • Current State: Pending work and next steps

Output Format

Markdown (Default)

markdown
# Brain Briefing: ~/signal

**Generated**: 2026-02-06 09:30 UTC
**Scope**: ~/signal + 1 child directory
**Transcripts**: 12 sessions (5 in last 7 days)
**Tasks**: 23 total (18 completed, 3 in_progress, 2 pending)

---

## Executive Summary

Signal Collective analytics platform. Recent work focused on deploying
independent brain for client work, audio transcription reliability, and
brain communication infrastructure.

---

## Active Tasks (3)

### #21 [in_progress] Fix Klaviyo sync errors
- **Owner**: signal-collective-001
- **Started**: 2026-02-05 14:30 UTC
- **Session**: f753dbf8-08e4-4d22-9436-49d0ef7db46a
- **Blocked by**: None
- **Status**: Investigating root cause in signal-workers-v2

### #22 [pending] Update BigQuery schema
- **Blocked by**: #21
- **Priority**: High
- **Next**: Unblocks after Klaviyo sync fixed

### #23 [pending] Add Facebook Ads tracking
- **Blocked by**: None
- **Priority**: Medium
- **Ready to start**: Yes

---

## Recent Sessions (Last 7 Days)

### Session f753dbf8 (2026-02-06, 4h 15m)
**Branch**: feat/signal-brain
**Tasks Completed**: #20 (Deploy Signal Collective brain)
**Files Modified**: 8
- brain/config.json
- brain/CLAUDE.md
- brain/MEMORY.md
- memory/MEMORY.md

**Key Activities**:
- Deployed independent brain for Signal Collective
- Configured standalone mode (no Derivative connection)
- Updated capabilities: klaviyo-analytics, bigquery-sync, shopify-integration

**Decisions**:
- ✅ Chose standalone brain (no AgentDispatch hub for client work)
- ✅ Set reportsTo: null (independent operation)
- ✅ Agent ID: signal-collective-001

### Session c5fc2201 (2026-02-05, 3h 50m)
**Branch**: main
**Tasks Completed**: #18 (Audio transcription reliability)
**PRs**: #128 (merged)
**Files Modified**: 12

**Key Activities**:
- Added audio transcription endpoint to mech-llms
- Fixed multipart FormData issues with Together.ai Whisper API
- Tested with 5.9MB audio file (successful)

**Decisions**:
- ✅ Use native FormData + Blob (not npm form-data package)
- ✅ Let fetch auto-set Content-Type boundary
- ❌ Rejected: Buffer approach (Together.ai returned 500)

---

## Agent Team Coordination

**Team**: signal-refactor (3 members)
**Active**: 2026-02-04 10:15 - 2026-02-04 14:45 (4h 30m)

### Team Members:
1. **intent-refactor** (Owner: claude-sonnet-4.5)
   - Tasks: #15, #16
   - Focus: Refactor intent handling system

2. **sdk-cleanup** (Owner: claude-sonnet-4.5)
   - Tasks: #17
   - Focus: Claude Agent SDK integration

3. **websocket-refactor** (Owner: claude-sonnet-4.5)
   - Tasks: #18, #19
   - Focus: WebSocket connection reliability

### Inter-Agent Messages: 47
- intent-refactor → sdk-cleanup: 12 messages
- sdk-cleanup → websocket-refactor: 8 messages
- websocket-refactor → intent-refactor: 5 messages

### Outcome:
✅ All tasks completed, merged to main
✅ No file conflicts (good task scoping)
⚠️ Websocket-refactor required lead intervention (debugging session timeout)

---

## Key Decisions Log

### 2026-02-06: Independent Brain Architecture
**Context**: Deploying brain for Signal Collective client work
**Decision**: Configure as standalone (no Derivative portfolio connection)
**Rationale**: External client work should remain independent
**Alternatives Considered**:
- ❌ Join mech-services group (rejected: not Derivative infrastructure)
- ❌ Report to decisive-gm (rejected: separate business entity)

### 2026-02-05: Audio Transcription Implementation
**Context**: Audio uploads failing with multipart boundary errors
**Decision**: Use native FormData with Blob for file uploads
**Rationale**: Together.ai Whisper API requires proper multipart encoding
**Alternatives Considered**:
- ❌ npm form-data package (rejected: created wrong headers)
- ❌ Buffer approach (rejected: 500 error from Together.ai)

### 2026-02-05: AgentDispatch Hub Deployment
**Context**: Need persistent brain communication infrastructure
**Decision**: Deploy to Fly.io with Mech Storage persistence
**Rationale**: 24/7 uptime, survives local restarts
**Alternatives Considered**:
- ❌ Local hub process (rejected: not reliable for production)

---

## File Modification Frequency (Top 10)

| File | Modifications | Last Modified | Sessions |
|------|---------------|---------------|----------|
| brain/config.json | 12 | 2026-02-06 | 3 |
| memory/MEMORY.md | 8 | 2026-02-06 | 4 |
| signal-workers-v2/src/sync.ts | 7 | 2026-02-05 | 2 |
| brain/CLAUDE.md | 6 | 2026-02-06 | 2 |
| mech-llms/src/routes/audio.route.ts | 5 | 2026-02-05 | 1 |

---

## Child Directories

### ~/signal/signal-workers-v2
- **Transcripts**: 8 sessions
- **Focus**: Klaviyo analytics, BigQuery sync
- **Recent Work**: Fixing sync errors, schema updates
- **Active Tasks**: #21 (Klaviyo sync errors)

---

## Pending Work

### Ready to Start (1)
- #23: Add Facebook Ads tracking (no blockers)

### Blocked (1)
- #22: Update BigQuery schema (blocked by #21)

### In Progress (1)
- #21: Fix Klaviyo sync errors (owner: signal-collective-001)

---

## Recommendations

1. **Next Action**: Complete task #21 (Klaviyo sync) to unblock #22
2. **High Priority**: None blocked, all progressing normally
3. **Technical Debt**: Consider adding retry logic to BigQuery sync
4. **Knowledge Gap**: Document Klaviyo API rate limits for future reference

---

**Last Updated**: 2026-02-06 09:30 UTC
**Generated by**: brain-briefing v1.0

JSON Format

json
{
  "generated": "2026-02-06T09:30:00Z",
  "scope": {
    "directory": "~/signal",
    "recursive": true,
    "childDirectories": ["~/signal/signal-workers-v2"]
  },
  "summary": {
    "transcripts": 12,
    "recentTranscripts": 5,
    "daysCovered": 7,
    "totalTasks": 23,
    "completedTasks": 18,
    "activeTasks": 3,
    "pendingTasks": 2
  },
  "tasks": [
    {
      "id": "21",
      "status": "in_progress",
      "subject": "Fix Klaviyo sync errors",
      "owner": "signal-collective-001",
      "sessionId": "f753dbf8-08e4-4d22-9436-49d0ef7db46a",
      "blockedBy": [],
      "blocks": ["22"]
    }
  ],
  "sessions": [
    {
      "sessionId": "f753dbf8-08e4-4d22-9436-49d0ef7db46a",
      "startTime": "2026-02-06T05:15:00Z",
      "duration": "4h 15m",
      "branch": "feat/signal-brain",
      "tasksCompleted": ["20"],
      "filesModified": 8,
      "keyDecisions": [
        {
          "topic": "Independent Brain Architecture",
          "decision": "Configure as standalone",
          "rationale": "External client work should remain independent"
        }
      ]
    }
  ],
  "teams": [
    {
      "name": "signal-refactor",
      "active": "2026-02-04T10:15:00Z",
      "duration": "4h 30m",
      "members": ["intent-refactor", "sdk-cleanup", "websocket-refactor"],
      "messages": 47,
      "outcome": "All tasks completed"
    }
  ]
}

Integration with Other Systems

1. Autonomous Bootup

When a brain runs /autonomous-bootup, it should call /brain-briefing automatically:

bash
# In brain bootstrap process
/brain-briefing $(pwd) --last 30

2. Session Start Hook

Add to .claude/hooks/session_start.mjs:

javascript
import { exec } from 'child_process';
import { promisify } from 'util';

const execAsync = promisify(exec);

// Generate briefing on session start
const { stdout } = await execAsync(`claude-skill brain-briefing ${process.cwd()} --last 7`);
console.log(stdout);

3. Daily Reports

Schedule automated briefings via cron:

bash
# Daily briefing at 9am
0 9 * * * cd ~/signal && claude-skill brain-briefing . --last 1 >> memory/daily/$(date +%Y-%m-%d)-briefing.md

Architecture

Dependencies

  • transcript-query skill (uses TranscriptParser library)
  • Node.js fs/promises for task list reading
  • Claude Code task system (~/.claude/tasks/)
  • Agent teams system (~/.claude/teams/) (optional)

File Structure

code
.claude/skills/brain-briefing/
├── SKILL.md              # This file
├── brain-briefing.mjs    # CLI entry point
└── lib/
    ├── briefing-generator.js  # Core briefing logic
    ├── task-correlator.js     # Task-transcript correlation
    └── team-analyzer.js       # Agent team analysis

Limitations

  • Transcript parsing is CPU-intensive for large histories (cache results)
  • Task correlation requires tasks to be properly tagged with sessionId
  • Team analysis only works if agent teams were actually used
  • Large recursive directories may take time to process

Future Enhancements

  1. Vector Search: Use Mech Search for semantic queries across transcripts
  2. Auto-Summarization: Use Claude API to summarize long sessions
  3. Trend Analysis: Identify patterns over time (velocity, bug rates)
  4. Skill Extraction: Automatically identify reusable patterns
  5. Decision Graph: Build graph showing decision dependencies

Security & Privacy

  • Transcripts contain full conversation history including potentially sensitive data
  • Brain briefings should respect same privacy constraints as transcripts
  • Consider encrypting briefings if storing remotely
  • This skill is read-only (never modifies transcripts or tasks)

Created: 2026-02-06 Dependencies: transcript-query, Node.js 18+, Claude Code 2.1+ Related Skills: transcript-query, autonomous-bootup, memory-manager